Abstract: The hydraulic pump is a key component in hydraulic system. Monitoring its operation state and carrying out fault diagnosis are of important significance to the reliability of the whole hydraulic system. Based on the wavelet packet decomposition (WPD) and wavelet coefficients residual analysis method, the fault diagnosis approach by employing the outlet pressure of the pump is proposed for the condition monitoring of hydraulic pump. By analyzing the characteristic of the discharge pressure of the hydraulic pump, the wavelet packet analysis is applied to the pressure signal to extract the feature spectrum of the hydraulic pump. The relationships of the feature signals in different frequency bands corresponding to the different fault factors are established. The wavelet packet energy residuals are then used to diagnose the pump health status. The sensitivities of the fault diagnosis with different wavelet function are also discussed. For eliminating the border distortions in WPD, a twin sliding window scheme is created to improve the performance of the diagnosing system. The validating test results show that this WPD and residual analysis based method can effectively improve the accuracy of fault diagnosis in comparison to the conventional fast Fourier transform (FFT) method.

Key words: Fault diagnoses, Hydraulic pump, Residual analysis, Signal processing, Wavelet packet decomposition